rappidAI PROJECT STATUS
Current evidence, visible gaps and no silent upgrades.
A concise view of what is public for each model line, what remains incomplete and which documentation work comes next.
LAST REVIEWED
23 July 2026
f7eda1fb0ae153f0f9cc3477ead997cbdb462b39
Research statements are reviewed against this immutable lumen-quantum snapshot and the pinned Hugging Face model revisions linked from each record.
Inspect pinned snapshotquantum-1-pilot
Public 49.3M-parameter experimental F16 GGUF release.
No standardized benchmark or final run manifest. Pilot weights, GGUF artifacts and tokenizers are all rights reserved; no public reuse license is granted.
quantum-1.6-pilot
Public 49.3M-parameter F16 GGUF release with reported continued pretraining and held-out metrics.
No raw evaluation record or complete run log. Pilot weights, GGUF artifacts and tokenizers are all rights reserved; no public reuse license is granted.
quantum-1-echelon
Architecture preflight, tokenizer validation and Garden pipeline smoke evidence.
No production dataset, trained model, checkpoint, output or benchmark.
OPEN PUBLICATION ITEMS
Gaps remain part of the record.
Pilot final run manifests
Immutable records linking data, code, tokenizer, completed steps, checkpoints, evaluation and release files.
Raw evaluation artifacts
Versioned prompts, outputs, decoding settings, metric commands and failure labels.
Resource measurements
Inference memory and throughput plus training hardware, runtime, cost and energy use.
Visual-asset provenance
Creator, source, permission, transformation and trademark records for project images and brand marks.
Echelon production and model artifacts
Final corpus manifest, training logs, checkpoints, weights, outputs and evaluation.
DOCUMENTATION ROADMAP
Evidence before claims.
- 01
Maintain the published all-rights-reserved pilot-artifact decision; attach an explicit reuse license only if the policy changes.
- 02
Attach raw, versioned pilot evaluation artifacts before making broader capability claims.
- 03
Complete the Echelon production-data workflow before any dataset-total claim.
- 04
Measure local inference requirements and performance using a documented environment.
- 05
Keep every website claim tied to an immutable artifact or an explicit missing-evidence label.
Primary sources
DOCUMENTATION DIRECTORY
Continue through the evidence.
Every section keeps public evidence, configured targets and missing information visibly separate.